2.
Abstract
An Abstract submitted for the approval of the faculty of the College of
Urban Planning and Policy at the University of Illinois at Chicago in
partial fulfillment of the requirements for the degree of Master of
Urban Planning in Transportation
Creating a Transit Service Index: Creating a Frequency Based Quality
of Service Index Using the General Data Feed Specification
by Andrew Keller
March 2012
This research explores the creation of a transit supply index (TSI) based on
the frequency of transit service. The index is created using data supplied
by transit agencies through the General Transit Feed Specification (GTFS).
The GTFS data tables can be queried and filtered to examine the number of
planned stops for an individual transit stop based on route, time of day, day
of the week and direction.
The frequency score can be analyzed using GIS software to apply these
frequencies to the service areas surrounding each transit stop. These
frequency scores are applied to an underlying geographic zone (census
tract, traffic analysis zone) and weighted by the proportion of service
coverage offered with the zone of study. The resulting scores can be used
to measure against standard benchmarks of coverage contained in the
Transit Capacity and Quality of Service Manual, be used in conjunction
with a demand index, or used as a basis for comparison of service level.
The TSI can be calculated for individual transit agencies, but can also be
used for areas with multiple transit agencies and overlapping service
coverage areas.
The paper will describe the methodology to create the index and examine
the sources consulted when creating this measurement. The paper will
examine applications for the TSI using northeastern Illinois as its study
area. The paper will also describe the challenges in creating the index
while exploring its limitations as a level of service tool.
2
3.
a le o
T
b f Contents
1. Abstract..................................................................................................................2
2. Table of Contents……………………………………………………………………3
3. Introduction………………………………………………………………………......4
4. Sources…………………………………………………………………………………..4
a. Framework…………………………………………………………………….….5
b. Transit Capacity and Quality of Service Manual……………….….5
c. General Transit Feed Specification………………………………… ….6
5. Methodology…………………………………………………………………………6
a. Basic Framework……………………………………………………………...7
b. Service Buffers………………………………………………………………....9
c. Overlapping Buffers…………………………………………………………10
6. Analysis……………………………………………………………………………….12
a. Map of Chicago Region TSI for Morning Peak Period………….12
b. Comparison by County……………………………………………………..13
7. Conclusions………………………………………………………………………….14
d………………………………………………………………………….16 8. Works Cite
9. Appendix
3
5. service index based on frequency of service at CTA rail stations. The frequency score
represents the product of transit trips per week and percentage of population served, or
net persons served (NPS). The score is based on block level data, aggregated to the level of
the census tract, and then graded based on scores relative to other tracts. Frequency
scores for stations are based on manual counts, and coverage area is defined as block
centroids within a half mile of a station. The assignment's methodology and scoring system
are based on literature from the Transit Capacity and Quality of Service Manual, Part 5:
Quality of Service. (TRB, 2003)
Figure 1: Formula for average NPS
S
ource: (Kawamura, 2011)
Transit Capacity and Quality of Service Manual (TCQSM) The guidelines for availability
of service are based on frequencies and coverage area as recommended by the TCQSM.
The manual measures quality of service from the passengers perspective and assigns letter
grades based on average headways during the peak period of transit, usually between 8‐9
am. The grades shouldn't be read as an assessment of the transit agency, and not all
communities should expect "A" level service. For smaller communities a LOS grade "C"
may be an appropriate threshold for minimum service levels. For large transit agencies,
ten‐minute headways may not provide enough transit supply to meet demand, leading to
vehicle crowding and a lower overall quality of service, despite "A" level service based on
eadways. (TRB, 2003) h
The grades are based on the attractiveness of transit to passengers based on the average
time between transit vehicles along a route. Figure 1 shows the grading scale, along with
comments describing the perception of this level of service to a passenger.
5
9.
Figure 3: Sample Calculation of the TSI
Route Buses Service Buffer Area Total Zone Area Percentage of Coverage TSI Headway (minutes)
1 24 0.5 1 50% 12 20
2 12 1 1 100% 12
3 48 0.5 1 50% 24 10
Total 84 2 1 200% 48 5
Example Score for Time Period Between 6-10 am
Coverage ScoreFrequency
20
Issues, Modifications, and Versions While the general framework for the TSI is designed
to be as straightforward and simple as possible a number of issues arise when translating
the guidelines of the Transit Capacity and Quality of Service Manual guidelines to a regional
scoring system. Three main questions emerged to adapt the TCQSM standards to a regional
model, scored by geographic zone. These questions involve the correct size for service
buffers, how to score areas with overlapping buffers, and how to measure transit vehicle
requencies. f
Service Buffers The goal for service coverage measures is to create a service buffer for the
areas surrounding a station that best represents an area that matches the actual area of
usage for transit users. However, a regional model also demands a level of simplification in
order to limit the number of calculations required to measure large areas. For the models
presented in this analysis a quarter‐mile Euclidean buffer is placed around each bus stop
and a half‐mile Euclidean buffer is placed around rail stations. These buffers are based on
studies that have concluded that 75 to 80 percent of bus users walk an average of a quarter
mile, or about 5 minutes, for bus stops. Rail passengers are willing to walk roughly twice
hat amount for an average of a half mile. (Transportation Research Board, 2003) t
For many reasons the Euclidean buffer does not serve as the most accurate measure for
service coverage. The physical layout of the space surrounding a transit station plays a
large role how far pedestrians are willing to walk and how much ground can be covered
within the time they are willing to walk. A GIS‐based network analyzer could simulate a
more accurate 5‐minute walking buffer based on actual street distance walked compared
9
14. Figure 7: Total Number of TAZ with Grades for Each County
Level of Service Total
Cook DuPage Kane Lake McHenry Will
A 471 34 8 21 0 3 537
B 53 9 2 11 2 4 81
C 66 15 9 12 0 2 104
D 60 22 14 15 6 5 122
E 29 18 10 6 1 3 67
F 173 126 102 110 95 171 777
Total 852 224 145 175 104 188 1,688
Level of Service Based on Headways Total by County
County
Figure 8: Total Number of TAZ with Grades for Each County
Level of Service Total
Cook DuPage Kane Lake McHenry Will
A 55 15 6 12 0 2 32
B 6 4 1 6 2 2
C 8 7 6 7 0 1
D 7 10 10 9 6 3
E 3 8 7 3 1 2
F 20 56 70 63 91 91 46
Level of Service Based on Headways Percentage by County
County
5
6
7
4
While this analysis can be useful on its own, the numbers would be more meaningful if
analyzed in conjunction with a demand index, or against minimum coverage thresholds.
The score would also have more meaning if connected to a measure of accessibility which
would more accurately rate the attractiveness of transit based on both frequency and
ccess. a
Conclusions
The strength of the TSI is its ability to be calculated from public sources in a matter that is
automated and allows for frequent updates, for both accurate up to date measurements of
service, or as a regular process of analysis that can be used to track historical trends in
level of service. It can be easily adapted to different geographic zones and the ability to
weight coverage based on percentage of a zone. This is useful for lower‐density areas with
14